Processor overlaps may bring tension to Qualcomm-NXP

Article By : Junko Yoshida

Qualcomm is attempting to repurpose Snapdragon processors for in-vehicle infotainment systems, where NXP has the older i.MX6 parts.

« Previously: After NXP takeover, what's next for Qualcomm?

Mike Demler, senior analyst at The Linley Group, posits that little overlap or conflict exists between Qualcomm and NXP, largely because “automotive is a new market for Qualcomm.”

However, the new partners have “overlapping multi-core ARM designs,” as Krewell noted. How they will reconcile two diverging product roadmaps—especially in the automotive market—remains to be seen.

It is clear that the new entity must tread carefully and find a way to promptly resolve any internal tensions, when they arise.

[IHS Qualcomm revenue (cr)]
__Figure 1:__ *Fiscal year 2015 Qualcomm’s revenue (Source: Qualcomm)*

Chris Webber, vice president of Automotive at Strategy Analytics, said, “Because automotive is currently such a small part of Qualcomm revenues (circa 2%), but such a large part of NXP’s revenues (64%), any slowness (perceived or real) in automotive growth could bring about tensions in the combined entity.”

“The relatively limited Qualcomm executive management experience and understanding of Automotive, and how the much greater expertise at NXP is integrated could be key in managing any such tensions,” he told EE Times. “However, Qualcomm clearly sees the value of NXP’s no. 1 auto position, and has obviously strong expansion plans in automotive."

Automotive apps processors

One area of conflicts is in the application processors for the automotive market.

Qualcomm is “attempting to repurpose Snapdragon mobile processors for in-vehicle infotainment systems, where NXP has the older i.MX6 parts,” Demler observed.

This isn’t a full-blown conflict—at least, not yet, according to Demler. It’s because Qualcomm’s parts “aren’t in any production vehicles at this point.”

But each company’s product offering in the automotive apps processor segment comes with its own pros and cons.

Krewell laid out, “The i.MX 8 [follow-on to i.MX6] might be vulnerable for the cockpit, but it has stronger case for sensor fusions and ADAS.” Meanwhile, “Qualcomm Snapdragons have better GPUs and this could give it the edge with displays,” he added.

Both Krewell and Demler suspect that the superior multimedia features in Qualcomm’s Snapdragon chips might ultimately win the internal debate.

Demler speculated that “the product roadmap will undoubtedly shift to a Snapdragon architecture.” The downside is that those mobile chips lack automotive features such as CAN bus, he added.

NXP’s automotive team is expected to make deeper inroads into automotive with its upcoming i.MX8 chip. Although i.MX8 is still pre-production, “it is purpose-built for automotive,” said Demler.

In a recent product briefing, NXP stressed the safety and reliability inherently designed into i.MX8 architecture. NXP designers made sure that the new architecture comes with built-in hardware-based virtualisation and domain protection.

The idea of the software-defined digital cockpit is getting traction among automotive companies. Many chip vendors advocate the use of one processor that runs both the infotainment system and a digital gauge cluster. Obviously, NXP’s i.MX8 is one of the many apps processors addressing that segment.

But there’s one big difference between i.MX8 and competing SoCs in the digital cockpit, said Kyle Fox, NXP’s product manager for the i.MX 8 series applications processor, during a recent interview with EE Times. “We offer a level of safety protection that no other apps processors have been able to provide before.”

Noting that hypervisors aren't reliably safe, NXP designed the i.MX8 architecture so that every IP resource, from GPUs to serial ports, has resource protection, ownership rights and access permissions built into its hardware, he explained.

Today, Qualcomm doesn’t have a processor equivalent to NXP’s i.MX8. Demler concluded, “So I wouldn’t see any reason to kill [i.MX8].”

Complementary roles

Both Steve Mollenkopf, CEO of Qualcomm, and NXP CEO Clemmer emphasised the complementary roles both companies will play in the combined business.

An example is multi-core ARM designs, used in automotive and a variety of other product segments.

Although both companies have overlapping multi-core ARM designs, each has different target markets, said Krewell. “Qualcomm and NXP have a wearables processor, but Qualcomm has a better connectivity story and NXP has a broader choice of micro-controllers. Qualcomm just started an embedded strategy, and NXP is much more advanced in embedded.”

Even in the automotive market, each company offers separate technical strengths and marketing experiences.

NXP is expected to lead in infotainment, safety systems, body and networking, powertrain and chassis, secure access. After the acquisition by Qualcomm, NXP will add Qualcomm’s telematics and connectivity to its portfolio.

IHS’ Buksh described NXP’s strength in “software defined radio (SDR), NFC, cyber security, CMOS radar, processors, automotive functional safety,” while Qualcomm has advantages in “5G, artificial intelligence, V2X, wireless charging.”

FD-SOI vs. bulk CMOS

What’s far from clear is the choice of semiconductor process technology for the combined companies’ future automotive apps processors.

For example, Freescale—well before it got acquired by NXP—accumulated extensive experience in designing MCUs and apps processors using a low-power FD-SOI process.

NXP’s i.MX8 is fabricated in a 28nm FD-SOI process technology. Noting that FD-SOI is much more resilient to radiation-induced soft errors and bit flips compared to a bulk CMOS-based process technology, NXP’s automotive team told us that the choice of process technology makes its apps processors “safer by an order of magnitude,” making the i.MX8 an ideal choice for automotive applications.

In contrast, Qualcomm has zero experience with non-bulk CMOS process technology. How the company feels about FD-SOI is far from clear.

During the conference call, Qualcomm’s Mollenkopf talked up “Qualcomm’s leading SoC capabilities.” He believes it will mesh well with “NXP’s leading industry sales channels and positions in automotive, security and IoT.” But Mollenkopf never mentioned a specific process technology for its future SoCs.

Many analysts expect Qualcomm’s volume advantage with foundries to trickle down to the automotive market. After all, Qualcomm’s engine has always been smartphone apps processors manufactured by using a CMOS process technology.

IHS’ Buksh specifically pointed out, “On the cost perspective, the merged entity would allow to share cost of advanced technology nodes (<16nm), [while] sharing cost with various other industry segments.”

He explained, “Today traditional automotive suppliers are compelled to scale down the technology nodes because of high-performance processing at low power consumption. But the ROI is too low in automotive. Through the merger NXP can leverage Qualcomm’s ROI from other industries to invest in new mask sets.”

Strategy Analytics’ Webber isn’t so sure. “Ultimately customers will decide. Automotive applications are about cost, performance and power trade-offs–so does a new application design have high performance requirement justifying higher cost of 14nm FinFET process?” Webber explained that i.MX8 is on 28nm FD-SOI, while Snapdragon 820A on Samsung’s 14nm FinFET.

The Linley Group’s Demler had this to say. “Building chips to automotive environmental and reliability specifications is new for Qualcomm, so they should carefully evaluate NXP’s choice of FD-SOI for the automotive market.”

Next: How deep learning plays a role in Qualcomm-NXP deal

Leave a comment